LES Intercomparison of Drizzling Stratocumulus: DYCOMS-II RF02

Transcription

1 LES Intercomparison of Drizzling Stratocumulus: DYCOMS-II RF2 Andy Ackerman, NASA Ames Research Center Acknowledgments Magreet van Zanten, KNMI Bjorn Stevens, UCLA Markus Petters, CSU Participating Groups

2 Outline Motivation Case specifications Some results (ensemble, then group by group) time series profiles trends within ensemble Summary Questions and issues

3 Scientific Focus How do increasing numbers of submicron aerosol affect stratocumulus cloud cover liquid water path How does drizzle affect boundary layer dynamics entrainment bulk cloud properties How do predictions of drizzle in LES compare with observations? Does sedimentation of cloud droplets affect results? If so, is the response from different models consistent?

4 Results from Previous Workshop Case: DYCOMS-II RF1, with very dry inversion, droplet concentrations about 1 cm 3, and no precipitation below cloud base Most LES entrained overlying air faster than measurements indicated, resulting in a thin, cloud layer with LWP lower than observed Reduction of radiative cooling by thin clouds results in poorly mixed boundary layers negative feedback on further entrainment Limiting subgrid-scale mixing at inversion (ad hoc or by skill or luck of SGS model) reduces entrainment, resulting in well-mixed boundary layer with thick cloud layer

5 Drizzle and Entrainment in a Mixed Layer Model inversion entrainment drying Steady state moisture budget: sfc source = entrainment drying + precipitation sfc source precipitation Decreased drizzle leads to deeper boundary layer and thicker cloud (Pincus & Baker 1994) Considered a single meteorological scenario, with a moist inversion Whether entrainment deepens or thins a cloud layer depends on thermodynamic jumps at top of BL (Randall 1984)

6 Large-Eddy Simulations of Strongly Precipitating, Shallow, Stratocumulus-Topped Boundary Layers (Stevens et al. 1998) ASTEX case study (moist inversion) with CCN concentration of 25 cm 3, using bin microphysics and 2-stream radiative transfer Drizzle dries updrafts less evaporative cooling available to drive downdrafts Dry downdrafts cumuliform convection ( Bjerknes 1938) Moreover, light drizzle by reducing entrainment in PBLs with large jumps in moisture across the inversion might actually lessen entrainment drying thereby leading to deeper PBL clouds. Such scenarios are largely speculative and need to be considered further.

7 The Impact of Humidity above Stratiform Clouds on Indirect Aerosol Climate Forcing (Ackerman et al. 24) LES with bin microphysics and 2-stream radiative transfer based on three case studies: ASTEX (A29, 4th GCSS WG1 Workshop), FIRE-I (EUROCS intercomparison), and DYCOMS-II (RF1, 8th GCSS WG1 Workshop) Droplet sedimentation and drizzle consistently decrease with increasing numbers of sub-micron aerosol Entrainment consistently increases as water sedimentation decreases Response of LWP depends on humidity of air overlying boundary layer

8 Temperature and Moisture Jumps above Cloud Top 1% 8% 6% 4% ASTEX -2 q t (g/kg) -4-6 FIRE-I 2% -8 RF θ l (K)

9 Domain Averages 25 ASTEX Liquid Water Path (g m -2 ) FIRE-I dry ASTEX DYCOMS-II Droplet Concentration (cm -3 )

10 Response to Suppressing Water Sedimentation Altitude (m) ASTEX Precipitation (mm d -1 ) Liquid Water (g kg -1 ) Altitude (m) RF Precipitation (mm d -1 ) Liquid Water (g kg -1 )

11 Temperature and Moisture Jumps above Cloud Top -2 1% 8% 6% 4% ASTEX >35 cm -3 q t (g/kg) -4-6 RF2 FIRE-I 225 cm -3 2% -8 RF1 35 cm θ l (K)

12 Model Domain Wider than past GCSS stratocumulus domains to allow for larger scales of convective organization expected in drizzling regime: 6.4 x 6.4 x 1.5 km, x = y = 5 m, z = 5 m near surface and initial inversion Those able to use a stretched grid requested to use specified grid, with 96 layers

13 Initial Conditions and Forcings 1.2 u [m/s] v [m/s] thetal [K] 1. Altitude (km) qt [g/kg] ql [g/kg] rad_flx [W/m^2] 1. Altitude (km)

14 Initial Conditions and Forcings Radiation: Beer s Law parameterization from previous workshop, which includes heating at cloud base, cooling at and above cloud top (no hook for radiative term in droplet condensational growth equation) Subsidence: fixed divergence of horizontal wind (3.76 x 1 6 s 1 ) Coriolis: geostrophic wind profiles specified (by Bjorn) Surface fluxes: fix friction velocity at.28 m/s, surface Prandtl number at unity, surface temperature at 292 K, and 1% RH at surface (should be 98% because of salinity) Sponge: above 125 m with time constant of 1 s

15 Cloud Microphysics Leg averages of droplet number concentrations (N, cm 3 ) within cloudy air (defined by N > 2 cm 3 ): Fix N at 65 cm 3, if possible Flight Leg Open Cells Closed Cells Cloud Top 54 ± 14 6 ± 13 Cloud Base 56 ± 16 8 ± 17 If microphysics ignores sedimentation of cloud droplets, use integral over log-normal size distribution assuming Stokes sedimentation (v r 2 ): F = c(3/(4πn)) 2/3 q 5/3 l exp(5ln 2 σ g ) where c is taken from Rogers and Yau (1989) and σ g = 1.5 If unable to fix N, use idealized CCN spectrum based on measurements

16 Cloud Condensation Nuclei Within BL Above BL Using non-prognostic aerosol, cannot handle vertical variation in context of a BL that is deepening Dotted line is idealized bimodal fit for BL aerosol assuming ammonium bisulfate (lognormal, not a power law) Supersaturation for droplet activation specified to not to exceed 1% during first hour

17 Model Descriptions Group/Model Precipitation Cloud Droplet Team SGS Model Microphysics Sedimentation CSU/RAMS Deardorff 2 moment some Jiang CSU/SAM Deardorff Khairoutdinov yes Khairoutdinov and Kogan (2 moment) MetO Smag-Lilly 2 moment yes Lock MPI Deardorff 1 moment, no Chlond 2 moment NASA/DHARMA dynamic bin, yes Ackerman Smag-Lilly Wyant et al. (2 moment) NCAR Deardorff Wyant et al. no Moeng NRL/COAMPS Deardorff Khairoutdinov Golaz and Kogan U Redding/LEM Smag-Lilly 1 moment no Weinbrecht UCLA Savic-Jovcic, Stevens none U Utah Deardorff 1 moment? yes Zulauf, Krueger Utrecht-KNMI/DALES Deardorff none yes van Zanten, de Roode WVU Deardorff Khairoutdinov Lewellen w/ partial cloudiness and Kogan yes

18 Ensemble Requirements One simulation from each group w/ and w/o precipitation Precipitation must include warm rain or drizzle, not just cloud droplet sedimentation, and no sedimentation permitted in run w/o precipitation Specification must be followed for both simulations Nine groups satisfied these constraints: CSU (Khairoutdinov), MetO, MPI, NASA, NCAR, NRL, U Reading, U Utah, WVU Results from 13 groups shown here, just not included in ensemble

19 Ensemble Time Series 2 lwp (g/m^2) 95 zi (m) 1.4 wstar (m/s) precip (mm/d) 12 vhf (W/m^2) 24 shf (W/m^2) A bit low on LWP and high on entrainment Nowhere near enough drizzle, and vapor flux too large Drizzle decreases entrainment, convective velocity scale (integral of buoyancy flux), and surface vapor flux, but not LWP median

20 CSU (Jiang) 2 lwp (g/m^2) 2 lwp (g/m^2) zi (m) 95 zi (m) 9 zb (m) 9 zb (m) precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) Includes giant CCN, substantially suppressing droplet activation LWP nearly triples in response to light drizzle, and cloud cover increases 2

21 CSU (Khairoutdinov) 2 lwp (g/m^2) 2 lwp (g/m^2) qc sed no sed 5 4/cc 65/cc.4.2. qc sed no sed /cc 65/cc 95 zi (m) 95 zi (m) 9 zb (m) 9 zb (m) qc sed no sed 75 4/cc 65/cc qc sed no sed /cc 65/cc 1. precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) qc sed no sed.1 4/cc 65/cc 4 2 qc sed no sed 4 2 4/cc 65/cc LWP roughly doubles in response to cloud droplet sedimentation alone slightly decreases when drizzle is then included, and then increases when droplet concentrations reduced by 25%

22 MetO (Lock) 2 lwp (g/m^2) 2 lwp (g/m^2) monotone l_ fixed l_ var 5 monotone l_ fixed l_ var.4.2. monotone l_ fixed l_ var.4.2. monotone l_ fixed l_ var 95 zi (m) 95 zi (m) 9 zb (m) 9 zb (m) monotone l_ fixed l_ var 75 monotone l_ fixed l_ var monotone l_ fixed l_ var monotone l_ fixed l_ var 1. precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) monotone l_ fixed l_ var.1 monotone l_ fixed l_ var 4 2 monotone l_ fixed l_ var 4 2 monotone l_ fixed l_ var Variable mixing length in SGS model diminishes entrainment and doubles LWP; monotone advection of scalars furthers both trends

23 MPI (Chlond) 2 lwp (g/m^2) 2 lwp (g/m^2) Luepkes Kessler Luepkes Kessler 95 zi (m) 95 zi (m) 9 zb (m) 9 zb (m) Luepkes Kessler Luepkes Kessler 1. precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) Luepkes Kessler Luepkes Kessler Thick, overcast cloud is not maintained (w/ and w/o drizzle) Entrainment slows as radiative cooling diminishes One-parameter (Kessler) drizzle scheme has little effect; two-parameter scheme further diminishes LWP and cloud cover

24 NASA (Ackerman) 2 lwp (g/m^2) 2 lwp (g/m^2) Wyant+sed n65_125 n9_ Wyant+sed n65_125 n9_1 95 zi (m) 95 zi (m) 9 zb (m) 9 zb (m) Wyant+sed n65_125 n9_ Wyant+sed n65_125 n9_1 1. precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) Wyant+sed n65_125 n9_ Wyant+sed n65_125 n9_1 LWP increases (too much) with bin microphysics (lack radiative effect on droplet growth) Precipitation (brackets measurements when parameterized) reduces entrainment too much CCN in boundary layer not enough to maintain measured droplet concentration

25 NCAR (Moeng) 2 lwp (g/m^2) 2 lwp (g/m^2) zi (m) 95 zi (m) 9 zb (m) 9 zb (m) precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) Precipitation nearly as great as measured, substantially reduces LWP and cloud cover 2

26 NRL (Golaz) 2 lwp (g/m^2) 2 lwp (g/m^2) zi (m) 95 zi (m) 9 zb (m) 9 zb (m) precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) Precipitation reduces LWP Precipitating simulation is archetypical ensemble member 2 2

27 UCLA (Savic-Jovcic and Stevens) 2 lwp (g/m^2) 2 lwp (g/m^2) zi (m) 95 zi (m) 9 zb (m) 9 zb (m) precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) Precipitation limited to cloud droplet sedimentation, which increases entrainment and decreases LWP

28 U Reading (Weinbrecht) 2 lwp (g/m^2) 2 lwp (g/m^2) no backscat backscat no backscat backscat 95 zi (m) 95 zi (m) 9 zb (m) 9 zb (m) no backscat backscat no backscat backscat 1. precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) no backscat backscat 4 2 Precipitation has little effect Turning of stochastic backscatter (negative viscosity) increases LWP and cloud cover 4 2 no backscat backscat

29 U Utah (Zulauf and Krueger) 2 lwp (g/m^2) 2 lwp (g/m^2) w/o sed w/ sed w/o sed w/ sed 95 zi (m) 95 zi (m) 9 zb (m) 9 zb (m) w/o sed w/ sed w/o sed w/ sed 1. precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) w/o sed w/ sed 4 2 Precipitation w/o cloud droplet sedimentation has little effect Precipitation w/ cloud droplet sedimentation decreases entrainment and increases LWP 4 2 w/o sed w/ sed

30 Utrecht-KNMI (van Zanten and de Roode) 2 lwp (g/m^2) 2 lwp (g/m^2) zi (m) 95 zi (m) 9 zb (m) 9 zb (m) precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) Thick, overcast cloud is not maintained (w/ and w/o drizzle) Cloud droplet sedimentation (not drizzle) decreases LWP and cloud cover 2

31 WVU (Lewellen) 2 lwp (g/m^2) 2 lwp (g/m^2) w/o sed w/ sed w/o sed w/ sed 95 zi (m) 95 zi (m) 9 zb (m) 9 zb (m) w/o sed w/ sed w/o sed w/ sed 1. precip (mm/d) 1. precip (mm/d) 1 ndrop_cld (/cc) 1 ndrop_cld (/cc) w/o sed w/ sed 4 2 Precipitation w/o cloud droplet sedimentation has little effect Precipitation w/ cloud droplet sedimentation decreases entrainment and increases LWP 4 2 w/o sed w/ sed

32 CSU (Jiang) 1. precip (mm/d) 1. precip sd/mean 1 precip (max-mean)/sd Particularly narrow dispersion and range of precipitation 1

33 NCAR (Moeng) 1. precip (mm/d) 1. precip sd/mean 1 precip (max-mean)/sd Dispersion low Peak values more than 5 standard deviations from mean 1 Precipitation limited to very small area

34 Ensemble Profiles 1.2 u [m/s] v [m/s] thetal [K] qt [g/kg] 1. Altitude (km) ql [g/kg] qr [g/kg] ndrop_cld [cm^-3] 1. Altitude (km) precip [W/m^2] rad_flx [W/m^2] tot_tw [W/m^2] tot_qw [W/m^2] 1. Altitude (km) Geostrophic wind speeds too high, and total fluxes far from measurements Median precipitation remarkably similar to average in closed cells Mistakenly included an extra member in ensemble for these profiles, but precipitation-induced changes in total moisture flux seems inconsistent with other results, suggesting possible internal inconsistencies in ensemble member(s)

35 Ensemble Profiles 1.2 w_var [(m/s)^2] tke [m^2/s^2] thetal_var [K^2] qt_var [(g/kg)^2] 1..8 Altitude (km) w_skw [(m/s)^3] tot_boy [cm^2/s^3] tot_uw tot_vw 1..8 Altitude (km) Precipitation diminishes buoyancy flux and decreases w 2, and increases w 3 (away from observations) Precipitation diminishes buoyancy flux and decreases w 2, allow for more vigorous convection by decreasing entrainment through diminished surface fluxes and kinetic energy (?) Momentum flux disagreement suggests scales beyond extent of model domain

36 Response to Droplet Sedimentation 2 lwp (g/m^2) 2 lwp (g/m^2) 95 zi (m) 95 zi (m) CSU_Marat 5 w/ sed w/o sed 5 w/o sed lwp (g/m^2) 2 lwp (g/m^2) 95 zi (m) 95 zi (m) DHARMA 5 w/ sed w/o sed 5 w/ sed w/o sed lwp (g/m^2) 2 lwp (g/m^2) 95 zi (m) 95 zi (m) Utah 5 w/o sed 5 w/ sed w/o sed 75 75

37 Response to Droplet Sedimentation 2 lwp (g/m^2) 2 lwp (g/m^2) 95 zi (m) 95 zi (m) UCLA 5 w/ sed w/o sed lwp (g/m^2) 2 lwp (g/m^2) 95 zi (m) 95 zi (m) WVU 5 w/o sed 5 w/ sed w/o sed For all but UCLA, droplet sedimentation results in reduced entrainment and increased LWP, consistent with Ackerman et al. (24)

38 Trends within Ensemble Surface Precipitation (mm/d) LWP (g/m^2) Precipitation generally increases with LWP, as expected NCAR is exception to trend (LWP low and precipitation high) Should compare cloud base precipitation trend to H 3 N scaling found by Pawloska and Brenguier (23) and van Zanten and Stevens (25)

39 Trends within Ensemble 1.1 Entrainment Rate (cm/s) LWP (g/m^2) At low LWP, entrainment tends to increase with LWP (radiative cooling) Tendency reverses at higher LWP (entrainment drying) Should consider more sophisticated analysis along the lines done by Bjorn for previous workshop

40 Trends within Ensemble 1. Entrainment Rate (cm/s) Surface Precipitation (mm/d) Entrainment tends to decreases as precipitation increases

41 Trends within Ensemble 6 delta LWP (g/m^2) LWP (w/o precip) (g/m^2) The greater LWP is (well-mixed, radiatively driven stratocumulus), the more it tends to increase when precipitation is turned on X-axis was meant to be LWP w/o precipitation, but I mistakenly used LWP w/ precipitation instead

42 Summary Precipitation generally reduces w θ v, w 2, and entrainment, and increases w 3 Precipitation leads to increases in LWP and cloud cover in some, and decreases in other simulations; ensemble medians of both are unchanged Cloud droplet sedimentation generally decreases entrainment and increases LWP Tendencies within ensemble hold promise and require deeper thought and analysis Any robustness of tendencies should not be considered universal to stratocumulus, since response of BL dynamics and cloud properties to precipitation depends strongly on thermodynamic jumps above BL I am deeply grateful for the efforts of all the participants and those providing measurement analyses

43 Questions and Issues Fix geostrophic winds For models that don t fix droplet number, scale accumulation-mode number concentration to give average cloud droplet number concentration of 65 cm 3? While (if) changing the specification, might as well set RH at surface to 98% Any disagreement regarding 3-h averaging period? Should variations on grid stretching be permitted? If not, should we use WVU s grid above initial inversion? Assess significance of neglecting radiative term in droplet condensational growth